Sei sulla pagina 1di 21

!

!
"#$#%!&#'()*+!,*!-.&./0!#!1+/)%)2,3#%!#*)(#%+4 !

!
Bert Botma

!"#$%&'()*+&,(%#
In this paper I provide an analysis of the pattern of nasal harmony displayed in Yuhup, a Maku
language spoken in the Amazonas province of Brazil. While the available Yuhup data are extremely
limited, they bring to light an interesting property which seems to set the Yuhup pattern apart from the
more familiar patterns found in other Amazonian languages, such as Tuyuca, Tucano and Southern
Barasano. In these languages we find a complementary distribution between a series of nasals and a
series of voiced oral stops. In Tucano and Southern Barasano the latter are in some contexts realized as
prenasalized, though this nasalization has a different status than harmonic nasalization. Prenasalized
stops occur in non-harmonic contexts only and never trigger nasalization themselves. For this reason, it
has been suggested that the nasality of prenasalized stops in Tucano and Southern Barasano is
phonologically irrelevant (see e.g. Piggott 1992, Rice 1993, Piggott & van der Hulst 1997).
The nasal harmony pattern of Yuhup, like those of Tucano and Southern Barasano, is characterized
by a complementary distribution between nasals and nasal contours (i.e. prenasalized and postnasalized
stops). Nasals are found in nasalizing contexts; in non-nasalizing contexts prenasalized stops occur
initially and postnasalized stops finally. If Yuhup is like Tucano and Southern Barasano, then we
would not expect these nasal contours to be able to trigger nasal harmony. The problem, however, is
that they do: Yuhup has a number of suffixes which surface as nasalized when following a root-final
postnasalized stop. It is this nasalization that appears to make Yuhup typologically anomalous.
The aim of this paper is twofold. First, I will argue that the pattern of nasal harmony in languages
like Tuyuca and Southern Barasano receives a straightforward analysis if it is assumed that nasal
harmony operates at the level of the syllable. The second aim of this paper is to extend this analysis to
the harmony pattern of Yuhup. The account of Yuhup harmony that I present is speculative. However, I
hope to show that we can get a better insight into the Yuhup pattern if we assume, following
Government Phonology, that root-final consonants are analyzed as onsets of empty-headed syllables
(see e.g. Harris & Gussmann 2001).
This paper is organized as follows. In 2, I offer a typological overview of nasal harmony patterns,
with special emphasis on the type of nasal harmony that is found in Amazonian languages. Next, in 3,
I focus on two theoretical implications of this harmony pattern: the sonorant status of what appears to
be phonetically a series of voiced obstruent stops, and the allocation of harmonic nasalization to the
level of the syllable. The stage is then set for an analysis of the nasal harmony pattern of Yuhup, which
I present in 4. 5 concludes.

In Nancy Kula & Jeroen van de Weijer (eds.),


2.4 (2005), 1-21.

. Special issue of
http://www.lucl.leidenuniv.nl
ISSN 1574-4728

-.'&#-(&/0#

1"#2#&34(5(63#(7#%0805#90'/(%3#838&./8#
Nasal harmony can informally be defined as a process in which nasality surfaces not as a property of
just one segment but of a string of segments. Typological research on nasal harmony processes has
shown that there are a number of parameters according to which language-specific harmony patterns
may vary (see e.g. Schourup 1973, Piggott 1992 and Walker 1998). The following parameters can be
distinguished:
(1)

:0'0/.&.'8#(7#%0805#90'/(%3
a. Domain of nasalization
b. Trigger of nasalization
c. Direction of nasalization
d. Target range of nasalization
e. Behaviour of non-targets

Parameters (a)-(c) are self-evident. In (d), target range refers to the range of segment types that is
compatible with nasalization. In (e), non-targets are those segment types that are incompatible with
nasalization. Non-targets either block nasal harmony or are transparent to it. In the remainder of this
section I briefly discuss the parameters in (1), paying particular attention to (d) and (e).
To put the discussion on a concrete footing, consider first the pattern of nasal harmony that is
displayed in Warao, an isolate of Venezuela and Guyana (cf. Osborn 1966; see also Piggott 1992):
(2)

a. !!"#
"#"$"%"
!!"#$%
!&%!&'("

give it to him
summer
give them to him
shadow

b. )*+&&*
)*,&
-.
-

kind of bird
it broke
sun
walking

c. '!
!'%!,'

turtle
kind of tree

The forms in (2a) indicate that nasal harmony spreads rightwards to the end of the word, is initiated by
nasal consonants, targets vowels, laryngeals and /+ -/, and is blocked by other segment types. The
forms in (2b,c) indicate that Warao has underlyingly nasalized vowels; like nasals, these trigger
progressive nasalization.1 Note that in a nasal harmony system such as Warao the domain of nasal
harmony essentially depends on the number of adjacent nasalizable segments. Nasalization may be
limited to a single segment, as in [-"], or it may be a property of an entire word, as in [!!"#].
In other languages there is evidence to suggest that nasal harmony is bound by a prosodic unit, in
particular that of the syllable.2 Consider as an illustration the following forms from Secoya, a Western
Tucanoan language of Ecuador (cf. Johnson & Peeke 1962; see also Ploch 1999):

1
I have been unable to find any Warao forms with vowel sequences whose second, but not first, member is nasalized. Osborns
description suggests that such forms should be possible.
2
Piggott (1996) argues that nasal harmony in Kikongo is bound by the foot. In addition, it could be argued that the crosslinguistically frequent process of regressive tautosyllabic vowel nasalization constitutes a nasal harmony that is bound by the
rhyme.

3#

;0805#90'/(%3#,%#<*9*4#

(3)

a. &%!"
(.!!
!&"
#"/0'

variety of tree
armadillo
variety of ant
crayfish

b. )(!&'
-.("
"!

she is weaving
sweet potato
bread

c. $&$&
)&)&

variety of tree
arm band

The forms in (3a) show that Secoya, like Warao, has a progressive nasal harmony which is initiated by
nasals, targets vowels, laryngeals and /+ -/, and is blocked by other segment types. The forms in (3b)
show that Secoya, again like Warao, has underlyingly nasalized vowels. In (3c), however, we see an
important difference between Secoya and Warao. These forms indicate that when /+ -/ precede a
nasalized vowel, they surface as nasalized. This is unexpected, since the direction of nasalization in
Secoya is generally progressive. Hence, the forms in (3c) suggest that two separate nasal harmony
processes must be distinguished: a progressive harmony that is instantiated by nasals and nasalized
vowels, and a regressive, syllable-bound harmony that is instantiated by nasalized vowels and which
targets preceding onset sonorants. Following Piggott & Van der Hulst (1997), I will term this second
type of harmony syllable nasalization.3
It is important to note that the two harmony processes in Secoya operate independently of each
other. This becomes evident when we consider the two processes from a cross-linguistic perspective.
On the one hand, we saw that Warao has progressive nasal harmony but lacks syllable nasalization. On
the other hand, there are languages like Yoruba, a Benue-Congo language of Nigeria, which lack
progressive harmony but do have syllable nasalization. Dunstan (1964:163) gives the following
consonant inventory of Yoruba:4
(4)
45(8,=.#
7',+0&,=.#
%0805#
&04#
50&.'05#
65,).#

1
6
!

)
0

234

&

&3$

531

[#]
8
9
-

Yoruba has an underlying contrast between oral and nasalized vowels. When the non-nasal sonorants /9
8 - +/ precede a nasalized vowel, they are realized as [#*+:* )*$]. Other consonants fail to nasalize in this
environment, which suggests that Yoruba has a process of syllable nasalization that targets sonorants.
As we will see, syllable nasalization is also found in the pattern of nasal harmony that is displayed in
many Amazonian languages. As far as Yuhup is concerned, I will argue in 4 that syllable nasalization
is possible only if the syllable in question has a filled nucleus.
What unifies the nasal harmony patterns observed so far is that non-targets invariably block nasal
spread. However, there are also languages where non-targets are &'0%840'.%& to nasal spread. This
pattern is for instance displayed in Tuyuca, an Eastern Tucanoan language of Colombia, where we find
forms of the kind in (5) (unless noted otherwise, all Tuyuca data are taken from Walker 1998):

3
4

Ploch (1999) refers to it as nasal sharing.


Note in (4) that [ ] is not phonemic, but an allophone of / / before nasalized vowels.

(5)

-.'&#-(&/0#

a. $,"
%!&!$'#-

to illuminate
there
howler monkey
wind

b. !'$.
#').*
).;!
<!00',

badger
coal
Yapara rapids
bird
to kill

The forms in (5a) show that words containing sonorants and laryngeals only are nasalized throughout.
The forms in (5b) show that in Tuyuca nasalization is not always distributed across a contiguous string
of segments, since the nasal harmony process skips voiceless obstruents. This suggests that we are
dealing with a case of non-local nasalization: given a particular domain, nasalization is associated with
all nasalization targets, irrespective of the presence or position of any non-targets. This pattern of nasal
harmony appears to be restricted to South America, where it is found in a number of Tucanoan, Tupi,
Chibchan and Maku languages. In the remainder of this paper I will refer to languages that display this
harmony pattern as Tuyuca-type systems.
A number of generalizations can be made regarding Tuyuca-type systems. First, these systems
exhibit considerably less variation in the target range than harmony systems in which non-targets block
nasalization (see e.g. Walker 1998 for examples of the latter). In Tuyuca-type systems only voiceless
obstruents are transparent; all other segment types are predictably included in the harmonic target
range.
Second, Tuyuca-type systems display a complementary distribution between a series of voiced oral
stops and a series of nasals. The former occur in what may be termed oral words, the latter in nasal
words. This is shown by the Tuyuca (near-) minimal pairs in (6); note here that voiceless obstruents
occur in both oral and nasal words:
(6)

a. >'05#?(')8
1"$/
2")/*
0=5
'0
$*
0"

swollen
to lose
follow
bat
to bend
to tie

b. ;0805#?(')8
!'$.
#').
)(;<!0$&0
0',

badger
coal
Yapara rapids
bird
to prepare soup
to kill

The fact that nasals occur in nasal words only means that forms like *[!"$/] and *[#")/] are impossible.
Similarly, the fact that voiced oral stops occur in oral words only means that forms such as *[1'$.] and
*[2').] are impossible.
The complementary distribution of voiced oral stops and nasals suggests that the two have a single
underlying representation. Different interpretations have been offered as to the nature of these
segments. According to one approach, Tuyuca-type nasal harmony targets all voiced segment types,
nasalizing sonorants and turning voiced stops into nasals (see e.g. Pulleyblank 1989, Noske 1995,
Walker 1998).5 Alternatively, it has been argued that this type of harmony targets sonorants only (see
e.g. Piggott 1992, Rice 1993, Piggott & van der Hulst 1997); this view implies that the voiced oral
stops function as sonorants phonologically, parallel to other nasalization targets. The latter approach is
also adopted in Botma (2004), where it follows naturally from the claim that nasalization and voice are
in complementary distribution: nasalization is limited to sonorants, while voice is limited to obstruents;
I return to this issue in 3.1.
5
Pulleyblank (1989) assumes that nasalization is regulated by a nasal/voicing condition, which sanctions the feature
combination [+nasal,+voice] and rules out the combination [+nasal, voice]. The problem with this approach is that it is
essentially stipulative, since it is unclear just why nasalization and voice have this affinity.

;0805#90'/(%3#,%#<*9*4#

5#

A third property of Tuyuca-type systems is that in harmonic words it is impossible to determine the
direction of nasal spread. For this reason, it has been argued that in these systems nasalization is a
property of morphemes (see e.g. Piggott 1992, Noske 1995). However, not all Tuyuca-type systems
have as their harmonic domain entire morphemes. For instance, Southern Barasano has oral words, as
in (7a), nasal words, as in (7b), and partly nasal words, as in (7c). Piggott & van der Hulst (1997) refer
to the latter as disharmonic roots; note that in such roots the direction of nasalization +0% be
determined (data from Smith & Smith 1971):6
(7)

a. >'05#?(')8
+,".,' pathway
&.h
eye
*&,*
to feed
"0"
buttocks

b. ;0805#?(')8
&,!!&, rattle
!"0,
people
!,#!
none
$").
demon

c. :0'&53#%0805#?(')8
1'#!
mirror
,"!"
poison
,'!'!
woman
(""!!&!#! ten

The form [(""!!&!#!] in (7c) shows that nasality is lexically associated to a vowel, from which it
spreads rightwards, skipping any intervening obstruents. Based on forms of this type, Piggott & van der
Hulst argue that the harmonic domains in (7b,c) can be unified if it is assumed that Southern Barasano
has underlyingly nasalized vowels and a process of syllable nasalization. The latter assumption is
required to account for forms like [$").] and [,'!'!]: if nasalization spreads rightwards from the
leftmost nasalized vowel, then syllable nasalization must be invoked to ensure that any sonorant
preceding this vowel will also be nasalized. Tuyuca differs in this respect from Secoya: syllable
nasalization is optional in Secoya-type systems but required in Tuyuca-type systems.
In 3 I focus in more detail on the properties of the nasal harmony pattern in Tuyuca-type systems.
In 3.1 and 3.2 I argue that the segment type that underlies the alternating nasals and voiced oral stops
is that of a sonorant stop. In 3.3 I turn to the issue of syllable nasalization.

@"#2#&9.('.&,+05#,%&.'4'.&0&,(%#(7#A*3*+0B&34.#838&./8#
@"!"#A9.#8&0&*8#(7#=(,+.)#('05#+(%8(%0%&8#
One unifying property of Tuyuca-type nasal harmony systems is that the range of non-targets is limited
to voiceless obstruents. From this, two conclusions can be drawn: either the range of nasalizable
segment types is identified in terms of voicing, or it is identified in terms of sonorancy. The decision as
to which of these analyses is appropriate depends on the status of the voiced oral stops, which, as
noted, are in complementary distribution with nasals. While the phonetic realization of these stops
might be argued to reflect their obstruent status, there are good grounds to analyze them as sonorants
phonologically.
The first argument for this is typological. As Rice (1993) observes, most, perhaps all, languages
have a contrast between a series of obstruent stops and a series of sonorant stops. In most languages
these sonorant stops are phonetically realized as nasals. In Tuyuca-type systems, on the other hand,
nasals occur only in the context of a following nasalized vowel. This shows that in such systems
nasality is not an underlying property of sonorant stops, and thus that the oral variants of sonorant stops
should be regarded as underlying. If, by contrast, the oral variants were obstruents, these systems
would lack an underlying contrast between obstruent and sonorant stops a highly marked state of
affairs.
A second argument for according the voiced oral stops sonorant status is that it allows a unified
account of Tuyuca-type nasalization. In a feature-based account, this permits an analysis in which
[nasal] associates only to segments that are specified for [sonorant] (or, following Piggott (1992) and
Rice (1993), for Spontaneous Voice). These segments would then also include /1 2 5/. This
interpretation can be straightforwardly reanalyzed in terms of the Element-based Dependency

Disharmonic roots also occur in other Tucanoan languages; compare e.g. Noskes (1995) description of Tucano.

-.'&#-(&/0#

framework of Botma (2004), which combines insights from Element Theory (see Harris & Lindsey
1995) and Dependency Phonology (see especially Anderson & Ewen 1987).
A fundamental assumption of Element-based Dependency is that the phonetic interpretation of
elements depends on the structural position in which they occur. A case in point is the element |L|.
Generally speaking, |L| has the articulatory correlate of open approximation and the acoustic correlate
of periodicity. The specific interpretation of |L| depends on its position in the segmental structure. If |L|
occurs as a manner element (i.e. as head), as in (8a), the segment is identified as a sonorant. If |L|
occurs as a phonation element (i.e. as dependent), its interpretation is variable: it denotes nasalization if
there is also an |L| present in the head, as in (8b), and voice if there is no |L| present in the head, as in
(8c):
(8)

a.

b.
L

c.
|L| present L

|L| absent

In (9) I provide four concrete examples:


(9)

a.

N
|
L
|
A
/./

b.

N
|
L
|
A
/"/

c.
L

O
|
/
|
U
/$/

d.

O
|
/
|
U

/1/

(9a) represents the low vowel /./, which has a manner component consisting of |L| only. This |L| is
dominated by the nucleus constituent N, and itself dominates the place element |A|. (9b) represents
nasalized /"/, which is like /./ but has an additional dependent |L|. Since there is also an |L| present in
the head, this dependent |L| denotes nasalization. (9c) represents a voiceless labial plosive /$/, which
has a manner component that consists of the stop element |/|. This |/| is dominated by the onset
constituent O, and itself dominates the place element |U|. (9d), which represents /1/, has an additional
dependent |L|. This |L| denotes voice, given that there is no |L| present in the head. The contextsensitive interpretation of dependent |L| thus embodies the claim that nasalization and voice are in
complementary distribution.
In Element-based Dependency, voiced oral stops of the kind that are found in Tuyuca-type systems
are analyzed as sonorant stops. Sonorant stops have a complex manner component in which |L|
dominates |/|; they are therefore quite literally a combination of sonorancy and stopness. (10) represents
an onset sonorant stop with coronal place:
(10)

O
|
L
|
/
|
I

In Botma (2004) it is argued that the phonetic interpretation of sonorant stops is variable, ranging as it
does between a voiced oral consonant (either stopped or continuant), a nasal, and a nasal contour. This
variability is in part a matter of free variation, and is in part dependent on the phonological system of
the language concerned. As an example of the former, consider the following facts from Rotokas, a

7#

;0805#90'/(%3#,%#<*9*4#

Papuan language of New Guinea. Firchow & Firchow (1969) describe Rotokas as having two dialects,
Rotokas A and Rotokas B. Rotokas A contrasts voiceless stops and nasals. Rotokas B has a surface
contrast between voiceless stops and a series of consonants that is realized as voiced, with variable
continuancy and nasality. Firchow & Firchow observe that there does not appear to be a conditioning
factor for this variation.
(11)

C(&(D08#2#
$
)
!
#

C(&(D08#-#
$
)
1~>~!
2~,~9~#

&
;

&
5~?~;

Regarding the nasal realizations in Rotokas B, Firchow & Firchow (1969:274) report that they are
rarely heard except when a native speaker is trying to imitate a foreigners attempt at speaking
Rotokas.
In the UPSID database Rotokas is classified as lacking an underlying series of nasals (cf.
Maddieson 1984), a typologically marked state of affairs.7 Following Rice (1993), I take the position
that the phonetically variable consonants are underlyingly sonorants. More specifically, my claim is
that they function as sonorant stops. The Element-based Dependency representation of the labial series
is as in (12):
(12)

a.

C(&(D08#2#

b.

C(&(D08#-#

O
|
L
|
/
|
U

O
|
L
|
/
|
U

[!]

[1~>~!]

The realization of the sonorant stops as nasals in Rotokas A represents the unmarked state of affairs.
Most languages realize their sonorant stops as nasals, presumably because the acoustic signature of
nasals offers the perceptually most salient compromise of sonorancy and stopness. The variable
realization of the sonorant stops in Rotokas B is more marked. I consider this variablity a matter of
phonetic implementation; that is, the realizations do not form distinct phonological entities but
represent different phonetic options (for more examples of this kind, see Botma 2004).
In other languages, the phonetic realization of sonorant stops is determined by the phonological
system. This is the case in languages where nasals are in complementary distribution with voiced oral
consonants, as in Tuyuca-type systems. In such languages we are dealing with an underlying series of
bare sonorant stops. In harmonic contexts these are realized as nasals through association with
dependent |L|, as in (13):
(13)

O
|
L
|
/

7
97% of the languages in UPSID have a series of underlying nasals; note that Maddiesons data appear to have been taken from
the Rotokas B dialect.

-.'&#-(&/0#

Nasalization in Tuyuca-type systems can thus be interpreted as targeting all |L|-specified manner
components within the harmonic domain.8
In Tuyuca-type systems sonorant stops also have a variable phonetic interpretation. In some
Tuyuca-type systems, such as Tuyuca and Desano, the voiced oral consonants are realized as stops. In
other Tuyuca-type systems, such as Southern Barasano and Tucano, they are realized as nasal contours.
As an illustration, consider some oral words from Southern Barasano (cf. Piggott & van der Hulst
1997:95):
(14)

2",'
+. 1. ~ +.1.
1. 5' ~ 1.5'
). 1')" ~ ).1')"

grasshopper
come!
eater
grass

(*2",')
(*+" 1.)
(* 1" 5')
(*)" 1')")

The forms in (14) show that voiced oral stops are phonetically realized as prenasalized stops; this
prenasalization is required word-initially and optional word-internally. There are good grounds to
interpret prenasalization as being phonologically irrelevant. First, in word-internal position nasalization
is not a required property of voiced stops. Second, and more importantly, prenasalized stops never
trigger nasalization: a form like [+.: 1.] is impossible in Southern Barasano.
Different interpretations of the nasalization of voiced stops have been offered in the literature.
Noske (1995:153), for instance, observes that:
[in most Tucanoan languages]9 prenasalized stops occur only between a nasal and an oral
vowel, and sometimes in word-initial position. Prenasalization is therefore predictable by
rule, and should be accounted for by a spreading rule.
Piggott (1992:48) maintains that the variation between voiced stops and prenasals in languages like
Southern Barasano must be treated as a phonetic effect (see Rice 1993 for a similar view):
The nasal property of prenasalized stops is epiphenomenal; it is directly derivable
from the articulatory adjustments required to realize spontaneous voicing.
Both Noske and Piggott note that prenasalization is predictable; Noske accounts for it in terms of a
phonological process, while Piggott shifts the explanatory burden to the phonetic implemenation. The
difference between the two accounts is not entirely trivial: if it turns out that there are languages where
nasal contours in oral spans trigger nasalization themselves, then Piggotts analysis is in trouble, while
in Noskes account such nasalization can be ordered to apply after the creation of nasal contours. As we
will see in 4, Yuhup appears to be such a language.

@"1"#C((&B8*77,E#,%&.'0+&,(%8#,%#A*3*+0B&34.#838&./8#
Potential evidence against an analysis of voiced oral stops as sonorants in Tuyuca-type systems comes
from morphological alternations, particularly from patterns of suffixation. In common with other
Tucanoan languages, Tuyuca distinguishes two classes of suffixes. The first class comprises suffixes
that are realized as either oral or nasalized, depending on the presence of nasalization in the root. The
second class surfaces as either oral or nasalized, regardless of the presence of nasalization in the root.
Following Walker (1998), I refer to these two classes as alternating and fixed suffixes respectively,
where the latter class consists of a set of fixed oral and fixed nasal suffixes. Consider the following
examples (from Walker 1998:29-31; see also Barnes 1990:283-5):
8

I ignore here that the harmonic target range in Tuyuca-type systems also includes laryngeals; see Botma (2004:279-306) for
discussion of this issue.
9
Within the Tucanoan language family prenasalized stops occur in Northern Barasano, Siriono, Carapana, Cubeo and Piratapuyo
(cf. Noske 1995).

;0805#90'/(%3#,%#<*9*4#

(15)

a. 05&.'%0&,%6#8*77,E :
07&.'#('05#'((&
:
07&.'#%0805#'((& :

/-,"/
[)%)/2,"]
[%'.23']

imperative of warning
watch out or you will get scolded!
watch out or you will get burned!

b. 7,E.)#('05#8*77,E :
07&.'#('05#'((&
:
07&.'#%0805#'((& :

/-2./
[$=.224]
[)"&022.]

classifier for round objects


two-CL-flexible; two strings
beads-CL-flexible; necklace

c. 7,E.)#%0805#8*77,E :
07&.'#('05#'((&
:
07&.'#%0805#'((& :

/2!"/
[&'.2!"]
[")'2!"]

imperative of permission
allow me to dig
allow me to come

9#

(16) lists some examples of alternating suffixes (cf. Walker 1998:30):


(16)

25&.'%0&,%6#8*77,E.8
a.
b.
c.
d.
e.
f.
g.
h.
i.

-.
-(.
--.
-+=
-+'
-,"
-,*
-,'
-,.

animate plural
contrast
imperative
evidential
evidential
imperative of warning
specifier
adverbializer
plural nominative

(17) lists some examples of fixed suffixes (cf. Walker 1998:30):


(17)

F,E.)#('05##

a.
b.
c.
d.
e.
f.
g.
h.
i.

recent past
evidential
inanimate-SG-NOM
classifier
classifier
classifier
too much
classifier
classifier

-.
--.
-,"
-1.
-2.
-5.
-$"
-)'
-0.

F,E.)#%0805
j.
k.
l.
m.
n.
o.
p.
q.
r.

-%"
-$(
-3'
-!"
-#"
-;"
-$'
-)!
-0"

emphatic
singularizer
time(s)
classifier
at that instant
diminutive
classifier
classifier
continued action

The problem here, as Walker (1998) notes, is that the set of alternating suffixes does not contain any
forms that have initial voiceless or voiced stops. Following Barnes (1996), Walker takes this to mean
that both the voiceless and voiced stops function as obstruents, since this permits the generalization that
nasalization in suffixes targets sonorants only.
The Tuyuca facts are open to other interpretations, however. First of all, Walker observes that
voiced velar stops, in contrast to other voiced stops, )( occur in alternating suffixes. An example is the
verbal suffix /-5'/, which is realized as [5'] after oral roots and as [;!] after nasal roots. To account for
this, Walker (1998:30) follows Trigos (1988) analysis of a similar phenomenon in Tucano:
[In] Tucano, which exhibits the same suffixal blocking effects, the velar nasal
alternant is actually a placeless nasal segment, and thus belongs to a separate class from
the stops.
It is difficult to see, however, why an analysis of the nasal alternant as placeless provides any evidence
for the obstruent status of the oral alternant. What is more, the pattern of suffixation in Tucano appears

10

-.'&#-(&/0#

to be different from that in Tuyuca. For instance, in Tucano we find alternating verbal suffixes of the
kind in (18) (cf. West & Welch 1967; see also Piggott & Van der Hulst 1997:107):
(18)

a. /'(.2 1.

let me write

5'5"2!"

let me see

b. /'(.2+"

I wrote

5'5"2$'

I saw

c.

I ate
seeds

#'"2$=
5")#20*

I was
pretty thing

1./.2$=
/')*20*

As can be seen in (18a), the stop-initial suffix /-1./ (a cognate of Tuyuca /-!"/; cf. (15c)) has a nasal
and an oral allomorph, like /-+"/ in (18b). This suggests that underlyingly this segment is a sonorant
stop, which, like root-internal sonorant stops, surfaces as nasalized in harmonic spans. Initial voiceless
stops, as in (18c), are not targeted by nasal harmony.
Tucano also has a number of gender and classifier suffixes that have oral and nasal variants,
depending on the presence of nasalization in the root. Some examples are given in (19) (cf. Noske
1995:168ff.):10
(19)

[5= ~ ;(]
[5' ~ ;!]
[5. ~ ;"]
[5= ~ ;(]

animate-SG-MASC
animate-SG-FEM
classifier-round objects
classifier-long/branching objects

Tucano is similar to Tuyuca in that most alternating stop-initial suffixes have an initial velar (but not
all, as (18a) shows). One possible analysis would be to analyze the voiced velar stops as being
underlyingly placeless, in line with Trigo (1988). Note, though, that such an analysis requires an
explanation as to why placeless segments are more prone to nasalize than place-specified segments. I
will not pursue this matter any further here.11
Let us next consider an alternative explanation of the nasal harmony pattern in Tuyuca suffixes.
Following Piggott & Van der Hulst (1997), I argue that the asymmetry between roots and suffixes
results from two distinct harmony systems.12 With regard to Tucano, Piggott & Van der Hulst maintain
that root-internal nasalization is non-local: it targets sonorants and skips obstruents, the latter being
transparent to the harmony. In suffixes, on the other hand, nasalization is local: it spreads rightwards
from the root to the suffix until it is blocked by a non-nasalizable segment.
The forms in (18) indicate that the range of non-nasalizable segment types in Tucano suffixes is coextensive with that in roots. However, given that we are dealing with two separate nasalization
processes, there is no a priori reason why the harmonic target range in roots should parallel that in
suffixes. This, I propose, is the key to understanding the Tuyuca harmony pattern. Within Tuyuca roots
nasalization targets all sonorants. Within Tuyuca suffixes, however, the harmonic target range is more
limited, in that it includes vowels and approximants but not the relatively less sonorous sonorant stops.
This implies, then, that in suffixes sonorant stops pattern with obstruent stops. In Element-based
Dependency, the class of obstruent and sonorant stops can be identified by the presence of the manner
element |/| (cf. the representations in (9a) and (9b) above).
I conclude that the pattern of suffixation in Tuyuca does not constitute compelling evidence against
an analysis of voiced stops in Tuyuca-type systems as sonorants. The patterning of voiceless obstruents
and voiced stops should not be taken to imply that the latter are obstruents, too. An alternative analysis
is that nasalization targets all sonorants in roots but only a subset of sonorants (i.e. those lacking |/|) in

10
Other Tucanoan languages with alternating stop-initial suffixes include Tatuyo (Gomez-Imbert 1980), Southern Barasano
(Smith & Smith 1971) and Northern Barasano (Stolte & Stolte 1971).
11
Walker speculates that nasalization of / / may be due to the relative difficulty of maintaining voicing during a posterior oral
closure. This account follows It & Mesters (1997) analysis of [ ~ ] allophony in Tokyo Japanese.
12
Piggott & Van der Hulst do not discuss the issue of stop-initial alternating suffixes. Piggott (1992), on which much of Piggott
& Van der Hulsts account is based, notes that most fixed oral suffixes in Tucano begin with a voiceless consonant. This is
incorrect, with respect to Tucano and to Tucanoan languages in general.

;0805#90'/(%3#,%#<*9*4#

11#

suffixes.13 This analysis is based on the assumption that the nasal harmony pattern in roots is different
from that in suffixes. Note that this distinction is needed in any case to account for the fact that initial
voiceless stops in fixed suffixes, e.g. /-$"/ in (17g), are not transparent to nasalization.14

@"@"#G3550H5.#%0805,I0&,(%#
The second aspect that must be considered before tackling Yuhup concerns the role of the syllable in
Tuyuca-type harmony. Of primary interest is the observation, made in 2, that Tuyuca-type harmony
systems are characterized by obstruent transparency and syllable nasalization. Given that obstruent
transparency implies syllable nasalization (but not vice versa), the challenge is to provide an analysis
which relates these two characteristics.
The first issue that must be considered is the question whether syllable nasalization implies the
relevance of the unit of the syllable or whether it is merely a side-effect of a more general process of
nasalization. At first sight, there appear to be good grounds for taking the latter position. Given that the
interpretation of dependent |L| is determined by context, it might be suggested that Tuyuca-type
nasalization involves association of |L| to all |L|-headed manner types. A Southern Barasano word like
[$").] demon, as in (7a), would then have the surface representation in (20):
L
(20)
O
|
L
|
U

N
|
L
|
A

O
|
/
|
I

N
|
L
|
I

$*

"*

)*

.*

Here the fact that the first syllable of [$").] surfaces as nasalized is epiphenomenal: it is simply the
result of association of dependent |L| to |L|-headed manner types, and the first syllable happens to
consist of |L|-headed manner types only.
The analysis of Southern Barasano nasalization in (20) is conceptually similar to the featuregeometric account of Piggott (1992), where nasalization targets are identified by the presence of a
Spontaneous Voicing node, which provides the appropriate landing site for the feature [nasal]. In this
account, the form [+:.:).] involves association of [nasal] to all SV-specified segments. This is illustrated
in (21), where RT and SV are short for Root node and Spontaneous Voicing node:
(21)

$*
|
RT
|
SV

* "*
|
RT
|
SV

)*
|
RT

.*
|
RT
|
SV

[nasal]
This analysis accounts for obstruent transparency, while syllable nasalization is simply an automatic
consequence of the more general process of sonorant nasalization.15
13

The assumption that the pattern of nasalization in suffixes is sensitive to sonority would be in line with Walkers hierarchy of
segment nasalizability.
14
It is surprising that the constraints on segment nasalizability should be stricter in suffixes. A possible explanation is that most
Tuyuca suffixes appear to have been derived from lexical morphemes; see Barnes (1996) for some discussion of this issue.

12

-.'&#-(&/0#

However, closer inspection of the distribution of nasalization in Tuyuca-type systems reveals that
this account is problematic. In Southern Barasano, the problem becomes clear when we consider
disharmonic roots. Consider the forms in (22), repeated from (7c):
(22)

*'#!
,"!"
,'!'!
(""!!&!#!

mirror
poison
woman
ten

Disharmonic roots bring to light two key properties of the Southern Barasano pattern. First, they show
that nasalization is progressive, spreading from the nasalization trigger to the end of a word. Second,
they show that nasalization originates in the leftmost nasalized vowel rather than in the leftmost
nasalized segment. Note that there are two reasons for identifying the leftmost nasalized vowel as the
harmonic trigger: any sonorant preceding a nasalized vowel is predictably nasalized, as in [,"!"], and
some disharmonic roots have a vowel-initial harmonic domain, as in [(""!!&!#!].
The distribution of nasalization in disharmonic roots suggests two possible analyses of the Southern
Barasano pattern. One option would be to posit two separate nasalization processes, both triggered by
the same nasalized vowel: the first proceeds rightwards and is unbounded, targeting all nasalizable
segment types; the second proceeds leftwards, is syllable-bound, and targets immediately preceding
sonorants. The other option would be to recognize rightward nasalization only, and to attribute leftward
nasalization to the effect of syllable nasalization. The nasal harmony pattern of Southern Barasano, and
of Tuyuca-type systems in general, suggests that the latter approach is more appropriate.
A general argument for the relevance of syllable nasalization is that leftward harmony is restricted
to an immediately preceding nasalization target. This limited scope follows naturally if leftward nasal
harmony is syllable-bound. A second argument for syllable nasalization concerns the nature of
rightward nasal harmony in Southern Barasano. Being a Tuyuca-type system, Southern Barasano
harmony is characterized by obstruent transparency. As Piggott & Van der Hulst (1997) observe, the
transparency of obstruents is similar to the more general consonant transparency that is found in
processes of vowel harmony. Piggott & Van der Hulst (1997:88) propose that vowel harmony can be
expressed by the following schema, where ! represents a syllable, V the syllable nucleus, F the
association of the harmonic feature F with V, and " the direction of harmony. The phonetic
instantiation of F as a property of the targeted vowels is represented by f; the fact that f results
from association with harmonic F is signalled by means of coindexation:
(23)

Fi
|
|
(V) (V)
|
|
fi
fi

The key insight of this interpretation is that vowel harmony is defined at the level of the syllable or,
more precisely, at the level of syllable heads. Piggott & Van der Hulst (1997:97) note that, as a result,
consonant transparency is an automatic consequence of vowel harmony as a relation between syllable
heads.
Given the similarity between vowel harmony and nasal harmony, Piggott & Van der Hulst propose
that the harmony pattern of Tuyuca-type systems can also be expressed in terms of a relation between
syllable heads. However, matters are complicated by the fact that the range of consonant transparency
in Tuyuca-type harmony is more restricted than in vowel harmony, given that nasalization targets not
just vowels but all sonorants. Thus, the challenge for Piggott & Van der Hulst is to provide a relation

15
Piggott (1992), for independent reasons, assumes a process of SV-fusion, which merges all tautosyllabic SV-nodes. Given that
the SV-node dominates [nasal], SV-fusion thus implicitly recognizes nasalization at the level of the syllable. See Piggott (1992)
for details of this analysis.

;0805#90'/(%3#,%#<*9*4#

13#

between syllable-to-syllable nasal harmony, which nasalizes all vowels, and syllable-internal nasal
harmony (i.e. syllable nasalization), which nasalizes all tautosyllabic sonorants.
To account for syllable-internal nasal harmony, Piggott & Van der Hulst (1997:102) suggest that
making the feature [nasal] a property of the nucleus automatically makes [nasal] a property of all
nasalizable segments within the syllable:
It is a fundamental principle of linguistic structure that the properties of the head of a
construction are simultanously the properties of the entire construction. Consequently,
when [nasal] is associated with the head or nucleus of the syllable, it is automatically a
feature of the syllable itself. It should, therefore, be realized on all the segments in the
syllable that can be nasal-bearing.
It seems reasonable to suggest that here Piggott & van der Hulst are referring to features that are
specified at the level of the nucleus, and not to just any feature of the vowel that this nucleus
dominates.16 On this assumption, let us consider their hypothesis in relation to the Southern Barasano
form [$").]. In (24) harmonic nasalization, as represented by N, is underlyingly associated to the
head of the first syllable. " represents rightward spreading of nasalization at the level of the syllable.
Nasalized segments are identified by the presence of n, the phonetic instantiation of N:
(24)
(

Ni
|
C V )
| |
$* "* *
| |
ni ni

(
*

|
C V )
| |
)* .*
|
ni

This cannot be the complete story, however. Since Piggott & Van der Hulst allow for the possibility of
nasalized obstruents, an explanation is needed for the fact that the /)/ in (24) is not a suitable target for
nasalization. To this end they argue that [nasal] is located at a different position in sonorants and
obstruents: in sonorants [nasal] is a head feature, while in obstruents [nasal] is a dependent feature.
Piggott & Van der Hulst (1997:103) represent the difference as follows:
(25)

a.

b.

;0805,I.)#8(%('0%&#

RT !
|
[nasal]

;0805,I.)#(H8&'*.%&#

[nasal]

RT
|

Syllable nasalization can then be defined as association of [nasal] to all segment types in which [nasal]
can occur as head. Piggott & Van der Hulst formalize this in terms of a principle that they term
Consistency of Dependency Relations:
(26)

J(%8,8&.%+3#(7#K.4.%).%+3#C.50&,(%8#(JKC; cf. Piggott & van der Hulst 1997:104)


Every occurrence of an inherited feature must manifest the same dependency relation.

The combination of syllable nasalization and the principle in (26) accounts for obstruent transparency
in Tuyuca-type nasal harmony systems.

16

Another way of putting this is to say that a distinction is required between harmonic and non-harmonic features. An example of
the latter is the place element |A| in the English word
, which is a property of the nuclear vowel but not of any other segment
in the syllable. A reasonable assumption is that the (non-) harmonic nature of a feature depends on the prosodic level at which
this feature is specified.

14

-.'&#-(&/0#

Piggott & Van der Hulst do not discuss the internal structure of nasalized segments in any detail,
nor do they consider the relation between [nasal] and other features. For instance, it is unclear whether
there is any further motivation for the different locations of [nasal], or in what way [nasal] can be said
to be contrastive in obstruents. Nevertheless, the fundamental claim of Piggott & Van der Hulsts
approach that Tuyuca-type nasal harmony is syllable-based is attractive, and I will adopt it below.
It should be noted, though, that the Element-based Dependency interpretation of Tuyuca-type systems
differs from Piggott & Van der Hulsts in some minor but important respects.
In line with Piggott & Van der Hulst, I assume that in languages that display syllable nasalization,
nasality is underlyingly specified at the level of the nucleus. More precisely, my assumption is that a
syllable which is specified for harmonic nasality has the general structure in (27). Following Levin
(1985), I take the level of the syllable to be equivalent to the maximal projection of the syllable head N,
i.e. N. Harmonic nasality, as expressed by |L|, is a dependent of this node:
(27)

N
|
N
|
N
|

O
|

(27) states that in languages with syllable nasalization, nasality is never contrastive in a domain smaller
than the syllable.17
Consider next how the Element-based Dependency approach accounts for nasal words in a Tuyucatype harmony system. The representation of Southern Barasano [$").], derived from underlying /+")./,
is given in (28):
(28)

O
|
L
|
U
$*

N
|
N
|
!
N
|
5! L
|
A
*

"*

5!
!

5!

O
|
/
|
I

)*

N
|
N
|
N
|
L
|
I

5!

.*

In the surface form each of the nasalization targets is specified for dependent |L|. In line with Piggott &
Van der Hulst, these |L|s can be regarded as the phonetic instantiations of syllable-level nasality; this is
expressed in (28) by means of coindexation.
This analysis can be extended to disharmonic roots such as [,'!'!]. In this form the nucleus of the
penultimate syllable is underlyingly specified for dependent |L|. Syllable nasalization then associates |L|
to the vowel and the preceding labial sonorant stop, while rightward syllable-to-syllable nasalization
spreads harmonic |L| to the following nucleus.
(27) illustrates an important difference between Piggott & Van der Hulsts analysis and the present
approach. Piggott & Van der Hulst account for obstruent transparency in terms of the CDR: /)/ cannot
be nasalized, since the feature [nasal] here does not manifest the appropriate dependency relation. The
present approach is more restrictive, given that there is only one possible landing-site for nasalization:
17

This approach can be viewed as an extension of Kehreins (2002) prosodic theory of laryngeal contrasts to the level of the
syllable, in particular because in Element-based Dependency nasalization is treated on a structural par with laryngeal features
such as voice and aspiration.

15#

;0805#90'/(%3#,%#<*9*4#

the target of Tuyuca-type nasalization can be identified as the dependent position of any |L|-specified
manner component. In view of this, a modification of Piggott & Van der Hulsts CDR is in order. A
revised and generalized version is given in (29):
(29)

J(%8,8&.%+3#(7#K.4.%).%+3#C.50&,(%8#(revised version)
In a domain Xn, where Xn is a projection of X0 and Xn is specified for a harmonic element F, F is
a property of all structures that are of the same type as X0.

With respect to syllable nasalization, X0 stands for the nucleus, Xn for the syllable (i.e. the maximal
projection of the nucleus), and of the same type as X0 for containing the manner element |L|. An
important prediction of the revised version of the CDR is that syllable nasalization can trigger
nasalization only, not both nasalization and voicing. This seems to be borne out by cross-linguistic
data. Processes which nasalize sonorants and voice obstruents, which occur in for instance Navaho and
Maukak, are segment rather than syllable-based (see Botma 2004 for an analysis of such processes).
In the following section we will see that the pattern of nasal harmony that is displayed in Yuhup
provides further support for the syllable-level status of nasality in Tuyuca-type languages.

L"#A9.#<*9*4#+(%*%)'*/#
Yuhup is an endangered Maku language spoken by a seminomadic group of about 400 people,
distributed among ten small villages in the west of the Amazonas province of Brazil and across the
border in the Vaups region of Colombia. This is also the general area in which we find the Tucanoan
language family, which includes Tucano, Tuyuca and Southern Barasano. Lopes & Parker (1999) note
that many Yuhup speakers are fluent in Tucano, which is the emergent lingua franca in the Vaups
region.
Lopes & Parker (1999) is the only description of Yuhup that is generally available, and all data
below have been drawn from it. Unfortunately, the data cited by Lopes & Parker have been kept to an
absolute minimum [owing to] the sensitive political issues which arise in conjunction with studying
indigenous groups in Brazil (Lopes & Parker 1999:324).18 Indeed, the property which seems to make
Yuhup special is reflected by just a handful of forms.
Bearing these caveats in mind, consider first of all the forms in (30), which indicate that Yuhup
roots display sonorant nasalization and obstruent transparency:19
(30)

a. >'05#7('/8
$.@-(
+A(A@)
('@2
2'@5
)37"/"1

b. ;0805#7('/8
rock, stone
striped mullet
hole
species of fruit
foot

$"@)%
6%6@$
%!@#
#"@;
&6/6@!

paternal uncle
to sleep
to vomit
grease, fat, oil
large potato (M0%&9(8(/0)

The forms in (30) indicate that the target range of nasalization includes all sonorants and laryngeals,
while the range of non-targets is limited to supralaryngeal obstruents, which are transparent to the
harmonic process. Voiced oral stops, which are in complementary distribution with nasals, are realized
as prenasalized stops root-initially and as postnasalized stops root-finally.
According to Lopes & Parker, the domain of nasal harmony is that of the morpheme. This is
supported by forms of the kind in (31), which indicate that Yuhup compounds freely combine oral and
nasal roots:
18

Additional Yuhup data can be found in Del Vigna & Lopes (1987), Del Vigna (1991) and Lopes (1995), but I have not been
able to consult these sources.
19
The sequence [V V] (more precisely [
]) is the phonetic realization of an underlyingly laryngealized vowel. Lopes &
Parker note that such sequences behave as single nuclei phonologically.

16

(31)

-.'&#-(&/0#

>'05B>'05
>'05B;0805
;0805B>'05
;0805B;0805

)B2 1B@5
(.-2!7/7%
!##2)*@5
!'%2#"@+

eye
boa constrictor
species of palm ($',0'&.0#NE(''9,I0)
paca rodent

On the basis of such forms Lopes & Parker analyze nasality as a morpheme-level autosegment [nasal],
which docks with its corresponding morpheme and spreads iteratively, in a left-to-right fashion, to all
harmonic targets. This morphological domain can be reanalyzed in terms of the prosodic domain of the
syllable, however, since Lopes & Parker (1997:325) note that at the underlying level each morpheme
consists of one, and only one, syllable and, conversely, each syllable can be assumed to correspond to a
distinct morpheme. This suggests that Yuhup, in line with Tuyuca-type systems, displays syllable
nasalization. It should be noted, though, that my analysis of syllable structure differs from that of Lopes
& Parker. Specifically, my claim is that CVC roots are disyllabic, with the final C occupying the onset
position of an empty-headed syllable.
A general argument for recognizing final empty-headed syllables comes from an inspection of the
types of possible Yuhup roots. According to Lopes & Parker, Yuhup has the following canonical root
shapes:
(32)

C((&#8904.
CV
/&*/
CVC
/$A$/
VC
/A+/
CVCC
/).-)/
VCC
/%-#/

wing
larva of bot fly (K.'/0&(H,0#9(/,%,8)
iguana lizard
to kick
species of bat (K.8/()*8#'(&*%)*8)

On the basis of these root shapes Lopes & Parker contend that each Yuhup morpheme corresponds
precisely to a syllable. Given that the penultimate segment of a root which ends in two consonants is
predictably /-/, they conclude that the maximal syllable template in Yuhup is /CV-C/. But this is not the
only possible interpretation. Lopes & Parker (1999:331-2; emphasis mine) make the following
observation regarding long vowels, which occur in stressed syllables only:
Primary stress in Yuhup regularly falls on the final vowel/syllable of the word The
physical correlates of primary stress are greater amplitude, higher pitch, and 5.%6&9.%,%6
of the vowel Suffixes are treated as extrametrical since they are normally unstressed.
The following forms illustrate this process of metrical vowel lengthening:
(33)

a.
b.
c.
d.
e.

/)37%/
/)37B$/
/)*-)*-/
/j'-j'g/
/).-)/

[C)37%@]
[C)37B@$]
[)*C)*@-]
[8'C8'@9 ]
[C).@-)]

coati
clothes
species of snake
elbow
to kick

The forms in (33) suggest that metrical vowel lengthening takes place in three contexts: root-finally, as
in (33a), before a single root-final consonant, as in (33b-d), and before a cluster whose initial consonant
is /-/, as in (33e). In Lopes & Parkers model of syllable structure there is no shared property which
underlies these contexts. However, metrical vowel lengthening receives a unified treatment if two
assumptions are made: firstly, that a root-final consonant occupies an onset position supported by a
following empty nucleus, and, secondly, that preconsonantal /-/ occupies the dependent position of the
syllable nucleus.
There are a number of arguments that support this analysis. First of all, there seem to be no
restrictions on the final consonant of Yuhup roots. This would be unexpected if this consonant occupies

;0805#90'/(%3#,%#<*9*4#

17#

the syllable coda, since codas generally permit only a subset of the contrasts permitted by onsets.
Another argument comes from the distribution of /-/. The available Yuhup data indicate that /-/!can cooccur with practically any kind of preceding vowel. Based on the forms given by Lopes & Parker, six
of the nine Yuhup vowels occur with a following /-/.20 The remaining three vowels are /"/, /=/, and /'/. It
is quite possible that the lack of forms containing ['-] is accidental, nor is it unthinkable that the
absence of forms containing ["-] and [=-] is due to a co-occurrence restriction on a high unrounded
vowel and a following palatal glide. I speculate, then, that apart from this constraint there are no cooccurrence restrictions on nuclei containing /-/. If this is correct, the Yuhup syllable template is CV(V).
This, in turn, permits a unified account of the metrical lengthening process as targeting stressed open
syllables. Note that this makes the context of metrical lengthening in Yuhup essentially equivalent to
that in Icelandic, as analyzed by, among others, Botma (2001) and Harris & Gussmann (2001).
A third argument for an onset analysis of final consonants comes from the distribution of harmonic
nasality in root-suffix combinations. Lopes & Parker note that the majority of Yuhup suffixes are
alternating: they are realized as oral or nasal, depending on the presence of harmonic nasality in the
preceding root.21 This is shown by the pattern of allomorphy displayed by the progressive suffix /-"(/:
(34)

a.

working
getting married
shouting

1"@/2"(
2'/'@(2"(
7.@+2"(

b. #@!2'%*
$"@%2'%
6%6@$2'%

killing
hearing
sleeping

As was argued in 3.1, there are good grounds to analyze the nasality of contours as phonologically
irrelevant in Tuyuca-type systems. However, Yuhup appears to be different in this respect: strikingly,
the locative suffix surfaces as nasalized when following a root that ends in a postnasalized stop:
(35)

beating
drinking

)B@2 2'%
B@5 2'%

This is unexpected if the nasality of the postnasalized stop is phonologically inert. But if the nasality of
root-final postnasals is phonologically active, then the question is why this nasality does not surface as
a property of the entire word; that is, why are the forms in (35) not realized as *[):@#2'%] and *[:@;2'%]?
I would like to propose that the solution to this conundrum lies in the onset status of root-final
consonants. This analysis makes it possible to account for the nasalized suffix allomorph in terms of
syllable nasalization. Suppose that the final syllables of [)B@2 ] and [B@5 ] have the following structure
(; denotes an empty position; I ignore place specifications):
(36)

O
|
L
|
/

N
|
N
|
N
|
;*

20

Yuhup has a nine-vowel system consisting of /


/.
Lopes & Parker give two examples of fixed oral suffixes, [examples of fixed nasal suffixes.
21

] to be and [-

] plural. They do not provide any

18

-.'&#-(&/0#

In Yuhup, as in Tuyuca and Southern Barasano, harmonic nasality is allocated at the level of the
syllable, where it is underlyingly associated with the syllable head N. What makes Yuhup different is
that it allows word-final empty-headed syllables. While such syllables can be specified for harmonic
nasality, the revised CDR in (29) prohibits this nasality from being realized phonetically. The point is
that harmonic nasality can be attached only to targets of the same type as the head of the harmonic
domain. If the harmonic domain is headless, as in (36), no harmony can take place. Consequently, any
sonorant occupying the onset of an empty-headed syllable will be phonologically oral and phonetically
postnasalized. The result of this analysis is therefore that postnasalization in Yuhup can be relegated to
the phonetic implementation, similarly to what I suggested for other Tuyuca-type systems.
The account proposed here raises a number of questions. The first is why suffixation of /-"(/ to a
form like /)B@2 / results in a nasalized allomorph of the suffix. A reasonable answer is that /-"(/ is
incorporated in the root-final syllable. Since the presence of a vowel implies that this syllable is now
headed, harmonic nasalization is free to associate to all targets. But this account raises another, rather
more difficult question: if suffixation triggers syllable nasalization, then why is the root-final consonant
in [)B@2 2'%] realized as a postnasalized stop, and not as a nasal? Clearly, we cannot appeal to the
explanation that was suggested for Tuyuca in 3.2. In Tuyuca, sonorant stops fail to nasalize in suffixes
because the harmonic target range in suffixes is different from that in roots. But in the case at handthe
sonorant stop does not form part of the suffix; it only occupies the same syllable as the suffix.
What I suggest instead is that the absence of nasalization of the sonorant stop reflects another kind
of difference between roots and suffixes, a difference which can be clarified with the OptimalityTheoretic notion of faithfulness. If harmonic nasality is underlyingly associated to syllable heads, any
association of nasality to a segment will imply a faithfulness violation, since the output of such a
segment will have an additional dependent |L| as compared to its input. It has been proposed that
faithfulness violations are relative to the domain in which they occur (see e.g. McCarthy & Prince
1994). For instance, cross-linguistic evidence indicates that languages tend to value faithfulness in
roots more highly than faithfulness in affixes. Given this, the fact that nasalization fails to target the
postnasalized stops in [)B@2 '%] and [B@5 '%] might be taken to suggest that Yuhup is one such language:
syllable nasalization targets the suffix but not the root, because faithfulness to segmental inputs is more
important in roots than it is in suffixes.
Unfortunately, the forms in (35) are the only examples of anomalous nasalization provided by
Lopes & Parker. However, there is further support for the exceptional status of the postnasalized stops.
Consider the pattern of suffixation displayed by the locative suffix:
(37)

ROOT
a. )"+
b. )"!
c. -%2

ROOT+LOC
path
village
clothes

)"@+2+")
)"!2!")
-%2 2##)

on the path
in the village
on the clothes

Lopes & Parker take the underlying form of the locative suffix to be /-CV)/, where C and V are copies
of the two root-final segments.22 For present purposes the form in (37c) is relevant, since it shows that
the copy of a postnasal is realized not as a postnasal but as a plain nasal, with the following vowel copy
surfacing as nasalized. This, too, is surprising if the nasality of the postnasal is a matter of phonetic
implementation. I suggest that the final onset view can also account for this suffixation pattern. On the
assumption that /-%2 / contains a final nasalized empty-headed syllable, the revised CDR ensures that
this nasalization cannot be realized in the absence of a nuclear vowel. This vowel is supplied by
locative suffixation, which creates an appropriate context for harmonic nasalization.
The final question that must be addressed is perhaps the most challenging. If the above account is to
be accepted, an explanation must also be provided for the fact that empty-headed nasalized syllables
always appear to have a sonorant stop in their onset. In other words, why is it that we never seem to
find a nasalized suffix allomorph after (say) a root-final voiceless stop, as in the hypothetical /+")2"(/ "
*[+")'%]? This is an important question, to which I can only give a tentative answer. Suppose that the
22

For an Optimality-Theoretic analysis of this process, see Walker (2002).

;0805#90'/(%3#,%#<*9*4#

19#

nasal contours derive historically from plain nasals, and that these plain nasals consisted of sonorant
stop structures with a dependent |L|. The historical change that then took place was that this |L| was
delinked from the nasal, turning it into a bare sonorant stop, and promoted to the level of the syllable
head. This scenario is illustrated for final postnasals in (38):
(38)

O
|
L
|
/

!
5!

N
|
N
|
N
|
;*

!
>

O
|
L
|
/

N
|
N
|
N
|
;*

5!

This diachronic development thus involved a change from segmental to syllabic nasality. Perhaps it
was this development that is responsible for the syllable nasalization that we find in Tuyuca-type
languages today. It is clear, though, that this can be established only on the basis of solid historical and
comparative evidence a matter that I will leave for further research.

O"#J(%+5*8,(%#
In this paper I have presented a dependency-based approach to the nasal harmony pattern found in
Tuyuca-type languages, such as Tuyuca and Southern Barasano. I have argued that this harmony
pattern receives a straightforward interpretation if two assumptions are made. First, Tuyuca-type
languages have an underlying series of sonorant stops, which are realized as nasals in harmonic
domains and as voiced oral stops or nasal contours in non-harmonic domains. Second, Tuyuca-type
languages have syllable nasalization; underlyingly, harmonic nasality is a property of syllable heads,
not of individual segments.
Yuhup is a Tuyuca-type language, though one with a twist. The voiced oral stops of Yuhup are
phonetically realized as postnasalized stops root-finally. The twist is that this postnasalization, contrary
to what is expected, triggers nasalization of a following suffix. This would seem to suggest that these
stops are specified for nasality, even though they occur in contexts that are otherwise non-nasal. I have
argued that the nasalizing potential of postnasalized stops becomes less of a mystery if harmonic
nasality is allocated at the level of the syllable. This permits an explanation for why nasal harmony is
instantiated when a suffix is added: the incorporation of a suffix vowel into an underlyingly nasalized
syllable makes the syllable headed, and thereby creates a context in which nasal harmony can apply.
While the advantage of this analysis is that it brings Yuhup in line with the more familiar nasal
harmony patterns of Tuyuca and Southern Barasano, more data from Yuhup and from genetically
related languages is required to substantiate it.

C.7.'.%+.8#
Anderson, J. & C. Ewen (1987). :',%+,45.8# (7# K.4.%).%+3# :9(%(5(63"# Cambridge: Cambridge
University Press.
Barnes, J. (1990). Classifiers in Tuyuca. In D. Payne, ed., 2/0I(%,0%#P,%6*,8&,+8Q#G&*),.8#,%#P(?50%)#
G(*&9#2/.',+0%#P0%6*06.8. Austin: University of Texas Press. 273-292.
Barnes, J. (1996). Autosegments with three-way contrasts in Tuyuca. $%&.'%0&,(%05# R(*'%05# (7#
2/.',+0%#P,%6*,8&,+8 62. 31-58.
Botma, B. (2004). :9(%(5(6,+05#284.+&8#(7#;0805,&3Q#0%#N5./.%&BH08.)#K.4.%).%+3#2++(*%&. Doctoral
dissertation, University of Amsterdam.

20

-.'&#-(&/0#

Dunstan, E. (ed.) (1969). A?.5=.#;,6.',0%#P0%6*06.8. New York: Africana Publishing Corporation.


Firchow, I. & J. Firchow (1969). An abbreviated phoneme inventory. 2%&9'(4(5(6,+05#P,%6*,8&,+8 11.
271-276.
Gmez-Imbert, E. (1980). La nasalit en Tatuyo: phonologie ou morphologie? 2/.',%),0 5. 65-81.
Harris, J. & E. Gussmann (2001). Word-final onsets. Ms., University College London and University
of Gdansk. [ROA-575].
Harris, J. & G. Lindsey (1995). The elements of phonological representation. In J. Durand & F.
Katamba, eds., F'(%&,.'8#(7#:9(%(5(63. London/New York: Longman. 34-79.
It, J. & A. Mester (1997). Correspondence and compositionality: the Ga-!" # $%&'%(')*# '*# +%,%*-.-#
Phonology. In I. Roca, ed., K.',=0&,(%8#0%)#J(%8&'0,%&8#,%#:9(%(5(63. Oxford: Oxford University
Press. 419-462.
Johnson, O. & C. Peeke (1962). Phonemic units in the Secoya word. In B. Elson, ed., N+*0)(',0%#
$%),0%#P0%6*06.8#$. Norman, Oklahoma: SIL. 78-95.
Kehrein, W. (2002). :9(%(5(6,+05# C.4'.8.%&0&,(%# 0%)# :9(%.&,+# :908,%6Q# 277',+0&.8# 0%)# P0'3%6.058.
Tbingen: Niemeyer.
Levin, B. (1985). 2# S.&',+05# A9.('3# (7# G3550H,+,&3. Doctoral dissertation, Massachusetts Institute of
Technology.
Lopes, A. (1995). Fonologia da lingual Yuhup: uma abordagem no-linear. Masters thesis,
Universidade Federal de Santa Catarina, Brazil.
Lopes, A. & S. Parker (1999). Aspects of Yuhup phonology. $%&.'%0&,(%05# R(*'%05# (7# 2/.',+0%#
P,%6*,8&,+8 65. 324-342.
Maddieson, I. (1984). :0&&.'%8#(7#G(*%)8. Cambridge: Cambridge University Press.
Noske, M. (1995). A9.# A.'%0'3# T8.# (7# K,8&,%+&,=.# F.0&*'.8. Doctoral dissertation, University of
Chicago.
Osborn, H. (1966). Warao I: phonology and morphophonemics. $%&.'%0&,(%05# R(*'%05# (7# 2/.',+0%#
P,%6*,8&,+8 32. 108-132.
Piggott, G. (1992). Variability in feature dependency: the case of nasality. ;0&*'05# P0%6*06.# 0%)#
P,%6*,8&,+#A9.('3 10. 33-78.
Piggott, G. (1996). Implications of consonant nasalization for a theory of harmony. J0%0),0%#R(*'%05#
(7#P,%6*,8&,+8 41. 141-174.
Piggott, G. & H. van der Hulst (1997). Locality and the nature of nasal harmony. P,%6*0 103. 85-112.
Ploch, S. (1999). ;08058# (%# /3# S,%)Q# A9.# :9(%.&,+# 0%)# J(6%,&,=.# 244'(0+9# &(# &9.# :9(%(5(63# (7#
;0805,&3. Doctoral dissertation, SOAS, University of London.
Pulleyblank, D. (1989). Patterns of feature co-occurrence: the case of nasality. :'(+..),%68# (7# &9.#
2',I(%0#:9(%(5(63#J(%7.'.%+.#1#UJ(3(&.#:04.'8#VW. 98-115.
Rice, K. (1993). A reexamination of the feature [sonorant]: the status of sonorant obstruents.
P0%6*06. 69. 308-344.
Schourup, L. (1973). A cross-linguistic study of vowel nasalization. >9,(# G&0&.# X('D,%6# :04.'8# ,%#
P,%6*,8&,+8 15. 190-221.
Smith, R. & C. Smith (1971). Southern Barasano phonemics. P,%6*,8&,+8 78. 80-85.
Stolte, J. & N. Stolte (1971). A description of Northern Barasano phonology. P,%6*,8&,+8 78. 86-92.
Trigo, L. (1988). >%#&9.#:9(%(5(6,+05#K.',=0&,(%#0%)#-.90=,('#(7#;0805#Y5,).8. Doctoral dissertation,
Massachusetts Institute of Technology.
Vigna, D. del (1991). Segmentos complexos da lingual Yuhup. Masters thesis, Universidade de
Braslia.
Vigna, D. del & A. Lopes (1987). F(%(5(6,0#:'.5,/,%0'#)0#PZ%6*0#<*9*4. Brazil: Arquivo Lingstico,
ALEM.
Walker, R. (1998). ;0805,I0&,(%[# ;.*&'05# G.6/.%&8# 0%)# >40+,&3# N77.+&8. Doctoral dissertation,
University of California.
Walker, R. (2002). Yuhup prosodic morphology and a case of augmentation. ;NPG 32. 116-135.
West, B. & B. Welch (1967). The phonemic system of Tucano. In B. Elson, ed., :9(%./,+#838&./8#(7#
J(5*/H,0%#P0%6*06.8. Norman, Oklahoma: SIL. 11-24.

;0805#90'/(%3#,%#<*9*4#

Bert Botma
LUCL/Department of English
University of Leiden
P.N. van Eyckhof 4
P.O. Box 9515
2300 RA Leiden
The Netherlands
e.d.botma@let.leidenuniv.nl

21#

Potrebbero piacerti anche